On the Maximum Size of Proteins to Stay and Fold in the Cavity of GroEL underneath GroES*

GroEL encapsulates non-native protein in a folding cage underneath GroES (cis-cavity). Here we report the maximum size of the non-native protein to stay and fold in thecis-cavity. Using total soluble proteins ofEscherichia coli in denatured state as binding substrates and protease resistance as the measure of polypeptide held in thecis-cavity, it was estimated that thecis-cavity can accommodate up to ∼57-kDa non-native proteins. To know if a protein with nearly the maximum size can complete folding in the cis-cavity, we made a 54-kDa protein in which green fluorescent protein (GFP) and its blue fluorescent variant were fused tandem. This fusion protein was captured in the cis-cavity, and folding occurred there. Fluorescence resonance energy transfer proved that both GFP and blue fluorescent protein moieties of the same fused protein were able to fold into native structures in the cis-cavity. Consistently, simulated packing of crystal structures shows that two native GFPs just fit in the cis-cavity. A fusion protein of three GFPs (82 kDa) was also attempted, but, as expected, it was not captured in thecis-cavity.

GroEL encapsulates non-native protein in a folding cage underneath GroES (cis-cavity). Here we report the maximum size of the non-native protein to stay and fold in the cis-cavity. Using total soluble proteins of Escherichia coli in denatured state as binding substrates and protease resistance as the measure of polypeptide held in the cis-cavity, it was estimated that the cis-cavity can accommodate up to ϳ57-kDa non-native proteins. To know if a protein with nearly the maximum size can complete folding in the cis-cavity, we made a 54-kDa protein in which green fluorescent protein (GFP) and its blue fluorescent variant were fused tandem. This fusion protein was captured in the cis-cavity, and folding occurred there. Fluorescence resonance energy transfer proved that both GFP and blue fluorescent protein moieties of the same fused protein were able to fold into native structures in the cis-cavity. Consistently, simulated packing of crystal structures shows that two native GFPs just fit in the cis-cavity. A fusion protein of three GFPs (82 kDa) was also attempted, but, as expected, it was not captured in the cis-cavity.
Molecular chaperones are a ubiquitous and abundant group of proteins that play essential roles in folding, assembly, and translocation of other proteins within the cell (reviewed in Ref. 1). One remarkable class of such components is the chaperonins found in bacteria, chloroplasts, mitochondria, archaebacteria, and eukaryotic cytosol. The best studied of these chaperonins are GroEL and its co-chaperonin GroES from Escherichia coli (reviewed in Refs. [2][3][4]. GroEL is a large cylindrical protein complex comprising two heptameric rings of 57-kDa identical subunits, and these rings are stacked back to back (5). Electron microscopy (6) has indicated that the crystallographically disordered 23-amino acid C-terminal segments of the seven subunits appear to project into the central channel of the cylinder at the level of the ring-ring interface, and hence the central channel of GroEL would function as two cavities, one in each ring. GroES contains seven identical 10-kDa subunits assembled as one heptameric ring (7). The GroEL ring to which GroES is bound is referred to as cis-ring and the opposite ring as trans-ring.
Current understanding of the productive pathway of a GroEL reaction is as follows. (i) Non-native polypeptide binds to GroEL at near the inner rim of the central cavity. (ii) Binding of ATP to the polypeptide-containing ring of GroEL permits the binding of GroES to that ring accompanying the release of polypeptide into an enclosed cage, defined by the GroEL cavity and the dome of GroES, in which folding to the native state can proceed with aggregation being avoided (6, 8 -13). (iii) Bound ATP is hydrolyzed (8,14), and GroES is released upon subsequent ATP binding to the trans-ring of GroEL, permitting native or partially folded proteins to leave GroEL (15).
The mechanism described above imposes a physical limit on the size of a polypeptide whose GroEL-assisted folding is strictly GroES-dependent. Examination of the crystal structure of GroEL/GroES/ADP 7 complex suggests that GroES binding fixes a drastic upward and outward shift of the apical GroEL domains, thereby increasing the size of the central cavity and forming a dome-shaped chamber ϳ85 Å high and ϳ80 Å wide. This folding cage, termed the cis-cavity, has the volume (175,000 Å 3 ) equal to a globular protein of ϳ142 kDa, assuming a perfect fit (1.23 Å 3 /Da) (16), and this must be the theoretical size limit of protein to be contained in the cis-cavity. Actual size limit, however, appears to be far below this limit, since 75-kDa methylmalonyl-CoA mutase was not able to be held in the cis-cavity (9). This may be also the case for 72-kDa phage P22 tailspike protein, since its GroEL-mediated folding did not require GroES (17). These proteins in non-native state can bind to GroEL (in the absence of GroES) or to trans-ring of GroES/ GroEL (in the presence of GroES) but cannot fit in the ciscavity. The question then arises: what is the real size limit of an unfolded polypeptide that can be accommodated within the cis-cavity? To determine this, we have employed the total soluble protein of E. coli as binding substrate of GroEL. It was previously demonstrated that GroEL binds half of the denatured (in vitro) and 10 -15% of nascent (in vivo) soluble proteins of E. coli with molecular size ranging from about 10 to 150 kDa (18,19). However, cis-trans topology of the bound proteins was not clarified in these studies. Using limited proteolysis, we assessed the topology of polypeptides bound to GroEL.
Another question related to (but independent from) the above one is: how large a protein can fold within the cis-cavity? To our knowledge, the protomer of ribulose bisphosphate carboxylase from Rhodospirirum rubrum (55 kDa) is probably the largest molecule for which GroEL-mediated folding shows strict GroES dependence, an indication of folding in the ciscavity (20). However, ribulose bisphosphate carboxylase is a homodimer enzyme, and its catalytic activity, the measure of proper folding, can only be assayed after the release from GroEL to the bulk medium. As a consequence, it is not certain if the protomer of this enzyme completes the folding in the cis-cavity or after the escape from the cis-cavity. Completion of folding in the cis-cavity has been demonstrated only for three relatively small proteins, namely 33-kDa rhodanese (13), 27-kDa green fluorescent protein (GFP) 1 (13,21), and 20-kDa dihydrofolate reductase (11). Here, we generated fused dimer (54 kDa) and trimer (82 kDa) of GFP, and monitored their binding and folding in the cis-cavity. Our results from experiments using total E. coli soluble protein and a GFP dimer clearly show that up to 54 -57-kDa protein can be accommodated and can fold in the cis-cavity.
Construction of Plasmids for GFPs-All GFP mutants were generated by site-directed mutagenesis using the Kunkel method (22), and single-stranded DNA of plasmid was obtained by infecting E. coli CJ236 cells with helper phage M13KO7 (Amersham Pharmacia Biotech). A NheI-EcoRI fragment containing the gfp gene was isolated from wildtype GFP-expressing plasmid TU#58 (a kind gift from Dr. Martin Chalfie; Ref. 23), and inserted into pET21c (Novagen) in NheI-EcoRI sites (pET-wtGFP). Using single-stranded DNA of pET-wtGFP as a template, plasmid pET-GFPer was generated by site-directed mutagenesis. The plasmid pET-GFPer encodes a mutant GFP (F99S/M153T/ V163A) that shows improved folding efficiency (24). The peak wavelengths of the excitation and fluorescence spectra of this GFP mutant are identical to those of the wild-type GFP. Next, using single-stranded DNA of pET-GFPer as a template, pET-GFP (FRET) and pET-BFP (FRET) were generated. pET-GFP (FRET) encodes a GFP mutant that has two more amino acid substitutions, F64L/S65C, that allow the protein to have a single, red-shifted excitation peak. pET-BFP (FRET) encodes a mutant form of GFP in which three more amino acid substitutions, F64L/Y66H/Y145F, are introduced into pET-GFPer. This mutant emits blue light (25). In addition, both pET-GFP (FRET) and pET-BFP (FRET) contained mutations at the 3Ј-terminus of gfp gene in order to delete crystallographically disordered C-terminal six amino acids, which may have a negative effect on GFP fluorescence (26) and to generate a peptide linker segment connecting BFP to GFP in the fused protein to be made. The pET-GFP (FRET) was digested with XhoI, and both termini of the fragment containing gfp gene were ligated together. This procedure generated the plasmid pET-GFP (FRET) His, which expresses Cterminal histidine-tagged GFP. From the pET-BFP (FRET) , NheI-(NheI-compatible)XbaI fragment was isolated and ligated into NheI site of pET-GFP (FRET) His. The resultant plasmid, pET-BFPGFP(His), encoded a fusion protein of BFP and GFP with His tag at the C terminus. Amino acid sequence of the protein expressed from this plasmid was deduced as MA-{BFP(S2 . . . F64L/Y66H/F99S/Y145F/M153T/V163A . . . G232)}-SGGRLESGGS-{GFP(S2 . . . F64L/65C/F99S/M153T/V163A . . . G232)}-SGGRLEHHHHHH. Trypsin cleaves the peptide bonds after Arg in the linker region and in the C-terminal tail. The plasmid pET-BFP (FRET) His, which encoded C-terminal His-tagged BFP, was produced by inserting NspV-BssHII fragment from pET-BFP (FRET) into the similarly cut plasmid pET-GFP (FRET) His. For simplicity, these His-tagged mutant proteins are denoted by GFP, BFP, and BFP-GFP in this paper.
Biotinylation of E. coli Soluble Proteins-E. coli JM109 cells were cultured in 2ϫ YT broth and harvested by centrifugation. Cells were washed with 50 mM Tris-Cl (pH 7.5), and 50 mM NaCl, and sonicated in 20 mM Na-P i (pH 7.5), 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride, 10 g/ml DNase I, and 10 g/ml RNase A. The suspension was centrifuged (20,000 ϫ g for 20 min) and the supernatant was incubated with NHS-LC-Biotin (Pierce), a primary amine reactive biotinylation reagent, at a molar ratio to protein 75:1 assuming that the average M r of total E. coli soluble proteins was 35,000 (27). After a 1-h incubation at 25°C, the solution was dialyzed against 8 M urea.
Protease Sensitivity of GroEL-bound Polypeptides-All reactions were carried out at 25°C. Biotinylated E. coli soluble proteins in 8 M urea (125 M) was diluted 100-fold into the dilution buffer (50 mM HEPES-NaOH (pH 7.5), 10 mM Mg(CH 3 COO) 2 , 5 mM KCH 3 COO, and 5 mM DTT), which contained 2.5 M GroEL and 5.0 M GroES. The solution was incubated for 1 min, and nucleotide, 50 M ATP (1.43 mol/mol GroEL protomer) or 1 mM AMP-PNP (final concentrations), was added. In the case of ADP, the dilution buffer containing 2.5 M GroEL, hexokinase (0.45 units), 20 mM glucose, and 1 mM ADP, was preincubated for 5 min to remove contaminating ATP. To this solution, 125 M biotinylated E. coli soluble proteins in 8 M urea was diluted 100-fold and, after 1 min, 2.5 M GroES was added. The mixtures containing ATP, AMP-PNP, or ADP thus prepared were incubated for 5 min, and then proteinase K was added to 10 g/ml. After 30 min, reaction was stopped by addition of 5 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride. The reaction mixture was centrifuged briefly (17,000 ϫ g for 5 min at 25°C), and an aliquot (80 l) was applied to a gel filtration HPLC column (G3000SW XL , Tosoh). The column was equilibrated and eluted with 50 mM HEPES-NaOH (pH 7.0), 10 mM Mg(CH 3 COO) 2 , 5 mM KCH 3 COO, 100 mM Na 2 SO 4 at a flow rate of 0.5 ml/min. The (GroES/)polypeptide/GroEL complex was isolated, and was analyzed by SDS-PAGE followed by blotting to polyvinylidene difluoride membrane. Preparation of GFP Trimer-The sites of XhoI and XbaI were introduced into pET-GFPer at the 3Ј-terminus of the gfp gene and a resultant plasmid, pET-GFPerϩc, was digested with XhoI. The fragment that contained the gfp gene was ligated both ends together to attach 6ϫHis tag at the C terminus of GFP and was named pET-GFP(His). Two gfp genes containing NheI-XbaI fragment obtained from pET-GFPerϩc were ligated into the NheI site of pET-GFP(His), and a plasmid, pET-3GFP(His), encoding a GFP trimer with a C-terminal 6ϫHis tag was produced. The linker sequence connecting each GFP moiety was Arg-Gly-Leu. GFP trimers were expressed in E. coli BL21(DE3) transformed with pET-3GFP(His) as inclusion bodies. Inclusion bodies were washed with 1% Triton X-100, solubilized in 100 mM Na-P i (pH 7.5), 6 M guanidine-HCl, 5 mM 2-mercaptoethanol, and applied on a Ni-NTAagarose column equilibrated with the same buffer. The column was washed with 100 mM Na-P i (pH 7.5), 8 M urea, 5 mM 2-mercaptoethanol and was eluted with a linear gradient from 100 mM Na-P i (pH 7.5) to 100 mM sodium citrate (pH 4.0) in 8 M urea, and 5 mM 2-mercaptoethanol. Fraction of GFP trimer was dialyzed to 25 mM Tris-HCl (pH 7.5), 8 M urea, and stored at Ϫ80°C until use. This GFP trimer preparation was not completely homogeneous but pure enough for the following experiment. The protease sensitivity experiment was carried out in the same manner as described in the experiment with BFP-GFP, except that 20 M denatured GFP trimer in 25 mM Tris-HCl (pH 7.5), 8 M urea, 12.5 mM HCl, 1 mM DTT (final pH 3.0) was diluted 50-fold into the dilution buffer containing 1.0 M GroEL and 2.0 M GroES. Next, 20 M ATP and then 4 g/ml proteinase K were added as indicated.
Other Procedures-Concentrations of GroES and GFP were determined spectrophotometorically. Extinction coefficients at 280 nm used for GroES, BFP-GFP, BFP, and GFP were 1500 (28), 35,020, 16,170 and 18,850 M Ϫ1 cm Ϫ1 , respectively. The values for the GFPs were calculated based on the amino acid composition using DNASTAR software. Concentration of other proteins were determined with a method by Bradford (29). In this paper, the molar concentrations of GroEL, GroES are expressed as those of oligomers. SDS-PAGE was carried out using 13% of gels (30).

Size of the Proteins in the cis-Cavity-Polypeptides bound to
GroEL are highly susceptible to proteolysis (8,31,32). However, when GroEL binds polypeptide first and GroES and ATP (or ADP) next, undigested polypeptides remain (9 -11). This is because polypeptide attached at the cis-ring of GroEL becomes sequestered in the cis-cavity upon binding of GroES and is protected from protease. Taking advantage of this protection, we analyzed the size of the proteins that were able to be accommodated in the cis-cavity (Fig. 1). Total soluble protein of E. coli was urea-denatured, biotinylated, and used as binding substrate that was the mixture of proteins of various size (Fig.  1, Total labeled protein). The mixture was diluted into the buffer containing GroEL and GroES, and allowed to form polypeptide/GroEL complex. Then, for indicated cases, ATP, ADP, or AMP-PNP was added to form GroES/polypeptide/ GroEL complex. Note that the amount of ATP was slightly over the stoichiometric amount to the GroEL protomers (1.43 mol of ATP/mol of protomer) so that most ATP would be exhausted before the release of polypeptides from GroEL, which was also an ATP-requiring process. The complex was isolated with gel filtration HPLC, electrophoresed, blotted onto a membrane, and visualized by using biotin tag. Polypeptides of various size ranging from about 8 to 110 kDa were found to be associated with the GroEL complex for all samples (Fig. 1, lanes 1, 3, 5,  and 7). By treatment with proteinase K prior to HPLC, polypeptides associated with the complex that had been formed in the absence of nucleotide disappeared completely (Fig. 1,  lane 2). However, in the cases of the complexes that had been formed in the presence of nucleotides, significant amount of polypeptides were protected from digestion even though majority of bound polypeptides disappeared (Fig. 1, lanes 4, 6, and 8).
Polypeptides lost by the protease treatment may represent those bound to trans-ring of GroEL and possibly those improperly bound to cis-ring. The protease-resistant polypeptides represent a fraction of polypeptides held in the cis-cavity; noticeably, they are all smaller than the 57-kDa GroEL subunit. Therefore, we have concluded that the maximum molecular size to be contained in the cis-cavity is ϳ57 kDa.
BFP-GFP-The above experiments did not reveal whether folding in the cis-cavity would be possible for the proteins with nearly maximum size. To determine this, we made a 54 kDa BFP-GFP in which blue fluorescent variant of GFP (BFP) and green fluorescent GFP 2 were fused in a single polypeptide. Both BFP moiety and GFP moiety in the BFP-GFP purified from expressing E. coli cells were proved to have native structures from two criteria. As in the case of wild-type GFP, individual monomers of BFP and GFP were resistant to trypsin digestion ( Fig. 2A, left and center lanes). Shift of the original protein band to a slightly lower band with increasing incubation time reflected the cleavage of the C-terminal tail connecting His tag. When the BFP-GFP was treated with trypsin, the linker regions were cleaved by trypsin and the BFP-GFP band disappeared in parallel with appearance of stable bands of GFP and BFP that were no longer digested except for cleavage of the GFP C-terminal tail ( Fig. 2A, right lanes). Another support for native structures of BFP and GFP moieties in the BFP-GFP was the fluorescence resonance energy transfer (FRET). When the BFP moiety of the BFP-GFP was excited by 380 nm light, FIG. 3. Folding of the BFP-GFP in the cis-cavity. A, protection of the BFP-GFP from proteinase K digestion. Denatured BFP-GFP was diluted into the buffer containing GroEL and GroES in the absence (lanes 1 and 2) or presence (lanes 3 and 4) of nearly stoichiometric ATP and treated with proteinase K (lanes 2 and 4). The (GroES/)BFP-GFP/ GroEL complex were isolated, electrophoresed, blotted, and visualized with anti-GFP antibodies. B, gel filtration HPLC elution of the GroES/ BFP-GFP/GroEL complex formed in the presence of ATP. The BFP-GFP was diluted into the buffer containing GroEL and GroES in the presence of nearly stoichiometric ATP. After a 20-min incubation, the solution was applied to a G3000SW XL column. Elution was monitored with fluorescence at 510 nm excited by 380 nm, and UV absorbance at 280 nm. Scales of ordinates are arbitrary. The absorbance peaks at 17.5 and 15.5 min were bovine serum albumin and an unknown contaminating protein in it, respectively. C, fluorescence resonance energy transfer of the BFP-GFP held in the cis-cavity. Wavelength of excitation light was 380 nm. The GroES/BFP-GFP/GroEL complex formed in the presence of nearly stoichiometric ATP was isolated by gel filtration HPLC, and fluorescence spectrum was measured (shown by red). Fluorescence spectrum of 100% fully folded BFP-GFP at the same BFP-GFP concentration is calculated and shown (blue, 100%). Fluorescence spectrum of the 100% half-folded BFP-GFP (50% BFP-only molecule, 50% GFP-only molecules, 0% of fully folded BFP-GFP) is also shown for references (blue, 0%). Scale of ordinate is arbitrary. Details of the experiment are described under "Materials and Methods."  1 and 2) or presence (lanes 3 and 4) of nearly stoichiometric ATP. Then the solutions were treated with proteinase K (lanes 2 and 4). The (GroES/) BFP-GFP/GroEL complex was isolated, electrophoresed, blotted, and visualized with anti-GFP antibodies. Details of the experiment are described under "Materials and Methods." excitation of the GFP moiety was induced and eventually the BFP-GFP emitted green fluorescence at 520 nm (Fig. 2B, trace  1). This efficient FRET can occur only when BFP and GFP moieties are located very closely, in the same protein in this case. In fact, as BFP and GFP moieties in the same BFP-GFP were allowed to diffuse apart each other by trypsin cleavage of the linker region, FRET disappeared; the green fluorescence decreased, and the blue fluorescence from BFP increased (Fig.  2B, traces 2-6).
Folding of the BFP-GFP in the cis-Cavity-Experiments similar to those for E. coli soluble proteins were carried out for the BFP-GFP. The native BFP-GFP was denatured by acid and used for the experiments. Acid-denatured BFP-GFP was able to restore the native structure spontaneously upon dilution into the neutral pH buffer. Both BFP and GFP moieties in BFP-GFP folded in similar kinetics, and half-maximal fluorescence was recovered at 133 and 150 s after dilution, respectively (data not shown). These rates were several times slower than those of monomer BFP and GFP (21). Experimental procedures were dilution of acid denatured BFP-GFP into the neutral buffer containing GroEL and GroES, with or without addition of ATP, with or without proteinase K treatment, isolation of the GroEL complex with HPLC, electrophoresis, and immunoblotting. The denatured BFP-GFP was able to bind to GroEL in the absence of ATP (Fig. 3A, lane 1), but it was digested completely by proteinase K (Fig. 3A, lane 2). In contrast, when ATP was added, some amount of GroEL-bound BFP-GFP was resistant to proteolysis (Fig. 3A, lanes 3 and 4). This indicates that full length of the BFP-GFP polypeptide can be accommodated within the cis-cavity. Next, we monitored HPLC elution of the complex formed in the presence of ATP and found that the complex emitted green fluorescence of GFP by 380 nm excitation light, a suitable excitation light for BFP (Fig. 3B). This fluorescence was really emitted from the BFP-GFP held in the cis-cavity of the GroES/GroEL complex, since addition of EDTA, which induced the release of GroES from GroEL, resulted in disappearance of fluorescence from the GroEL peak fraction (data not shown). Folded, free BFP-GFP was also eluted at 19 min, which were probably BFP-GFP molecules released from the cis-cavity and/or trans-ring of GroEL. Because native BFP-GFP by itself does not bind to GroEL (data not shown), native BFP-GFP associated with the GroES/GroEL complex should be a product of folding in the cis-cavity.
To analyze contribution of FRET to the fluorescence at the GroEL peak fraction of HPLC, the GroES/BFP-GFP/GroEL complex was isolated, and its fluorescence spectrum was measured (Fig. 3C). In principle, the spectrum would have been the sum of contributions of fully folded BFP-GFP, in which both BFP and GFP moieties completed folding, and of half-folded BFP-GFP, in which only one of moieties completed folding. We estimated the relative contribution of the fully folded BFP-GFP in the observed spectrum on the assumption that the populations of two kinds of half-folded BFP-GFP ("BFP-only" molecule and "GFP-only" molecule) were the same. Based on this assumption and using the spectra in Fig. 2B as references, one can calculate that the spectrum in Fig. 3C is generated by a mixture of ϳ60% fully folded BFP-GFP and ϳ40% half-folded BFP-GFP. The spectrum of 100% fully folded BFP-GFP, as well as the spectrum of 100% half-folded BFP-GFP (ϭ0% fully folded BFP-GFP), at the same concentration of BFP-GFP are also calculated and shown in Fig. 3B (100% and 0%). The assumption of equal population of BFP-only molecules and GFP-only molecules may not be far from the case, since inherent folding ability of BFP and GFP moieties in BFP-GFP are similar as described in the above paragraph. Furthermore, even without this assumption, occurring of FRET in the GroES/ BFP-GFP/GroEL complex is evident. If whole fluorescence at 510 nm by 380 nm excitation light of the observed spectrum in Fig. 3C has been emitted solely by GFP-only molecules but not derived from FRET, we can predict the amount of GFP-only molecules and hence the intensity of the 510 nm fluorescence excited by 480 nm light, an optimum excitation light for GFP. However, the observed intensity at 510 nm excited by 480 nm was much smaller than the predicted one (data not shown) and could not be explained without FRET. Altogether, we conclude that significant fraction of BFP-GFP molecules within the ciscavity achieve complete folding; both BFP and GFP moieties reach the native structures.
Proteinase Digestion of GroEL-bound GFP Trimer-We also made an 82-kDa protein consisting of three GFPs. The denatured GFP trimer could bind to GroEL in the absence and presence of ATP (Fig. 4, lanes 1 and 3), but, as expected, bound GFP trimer was completely digested by proteinase K (Fig. 4,  lanes 2 and 4). ATP did not protect bound GFP trimer from digestion. The GFP trimer is too large to be accommodated in the cis-cavity. DISCUSSION Although the crystal structure of GroES/GroEL complex imposes ϳ142 kDa as a physical size limit of protein to be accommodated in the cis-cavity (16), the actual size limit has not been determined by experiments. Here, we have shown that up to ϳ57-kDa protein, but no larger proteins, can be accommodated in the cis-cavity and that at least a 54-kDa protein can finish folding there. Thus, actual size limit is less than half of the maximum physical size limit. Folding intermediates have more expanded volume, and random thermal motion of polypeptide chain of the intermediate should be allowed to find the proper FIG. 5. Simulated packing of two GFP molecules in the ciscavity. Simulation was carried out with Insight II computer program using crystal structure of GFP (33) and that of GroES/GroEL (16).
interactions between segments distant in the primary sequence. Also, because interactions with surrounding water molecules are essential to stabilize native structure of the soluble proteins, the cis-cavity must contain many water molecules for the folding intermediate to reach the native structure. Entropic burden to pack a large extended polypeptide chain into a restricted space may be another factor limiting the size.
The shapes of folded, native proteins are not always exactly globular, and the fit of the shape of native protein with that of the cis-cavity may also be important. The 54-kDa protein we tested for folding is a fused dimer of GFP, and a monomer GFP is a 27-kDa protein with the shape of a cylinder, which has a diameter of ϳ30 Å and length of ϳ40 Å (33,34). Using the crystal structures of GFP and GroES/GroEL complex, we tried to pack the GFP dimer into the cis-cavity and found that the dimer can just fit in the cis-cavity without steric collision (Fig.  5). Three GFP molecules cannot fit in by any means. Interestingly, a 56-kDa T4 phage head protein (gp23) may not be contained in the cis-cavity underneath GroES and the phage encodes its own derivative of GroES, gp31, which can form a larger cis-cavity to accommodate gp23 (35). Now that ϳ57 kDa is the cut-off size of the cis-cavity, how do the larger proteins fold in vivo? Pulse-labeling experiment of E. coli cells showed that proteins larger than 57 kDa actually bound to GroEL, even if the population was small and commitment of GroES was not clear (19). Furthermore, the bound large proteins were retained to GroEL in a more stable manner during the chase period than the proteins of molecular size 25-55 kDa. Implication of this observation remains open. It was recently reported that hsp60 (GroEL homologue)-mediated in vivo folding of a 87-kDa mitochondrial aconitase in yeast was dependent on hsp10 (GroES homologue) (36). The question of whether GroEL (and GroES) indeed assists the folding of large proteins in vivo has not been answered and should gain more attention.